Using Panel with PyVista

PyVista supports the usage of the panel module as a vtk.js jupyterlab plotting backend that can be utialized as either a standalone VTK viewer, or as a tightly integrated pyvista plotting backend. For example, within a Jupyter notebook environment, you can pass jupyter_backend='panel' to plot, or Plotter.show to automatically enable plotting with Juptyer and panel.

For example, here’s the PyVista logo:

from pyvista import demos
demos.plot_logo(background='white', jupyter_backend='panel')

Note that this isn’t a perfect replica since there are some details lost in the conversion to vtk.js, but for the vast majority of cases, this can be used to accurately render pyvista plots within Jupyterlab.

Examples and Usage

There are two ways to use panel within Jupyter notebooks. It can be done on a plot by plot basis by setting the jupyter_backend in mesh.plot():

import pyvista as pv
from pyvista import examples

# create a point cloud from lidar data and add height scalars
dataset = examples.download_lidar()
point_cloud = pv.PolyData(dataset.points[::100])
point_cloud['height'] = point_cloud.points[:, 2]
point_cloud.plot(window_size=[500, 500],
                 jupyter_backend='panel',
                 cmap='jet',
                 point_size=2,
                 background='w')

Alternatively, you can set the backend globally:

import math
import numpy
import numpy as np
import pyvista
from pyvista import examples

# set the global jupyterlab backend.  All plots from this point
# onward will use the ``panel`` backend and do not have to be
# specified in ``show``
pyvista.set_jupyter_backend('panel')
# create a sphere for Mars
sphere = pyvista.Sphere(
    radius=1, theta_resolution=90, phi_resolution=90, start_theta=270.001, end_theta=270
)
sphere.active_t_coords = numpy.zeros((sphere.points.shape[0], 2))
sphere.active_t_coords[:, 0] = 0.5 + np.arctan2(-sphere.points[:, 0], sphere.points[:, 1]) / (
    2 * math.pi
)
sphere.active_t_coords[:, 1] = 0.5 + np.arcsin(sphere.points[:, 2]) / math.pi

sphere.textures["surface"] = examples.planets.download_mars_surface(texture=True)

# with a black background
pl = pyvista.Plotter(window_size=[500, 500])
pl.background_color = 'black'
pl.add_mesh(sphere, smooth_shading=False)
pl.show()

Configuration Considerations

If running on a headless environment (e.g. Google Colab, your own VM), be sure to start up a virtual framebuffer using Xvfb. You can either start it using bash with:

export DISPLAY=:99.0
export PYVISTA_OFF_SCREEN=true
export PYVISTA_USE_IPYVTK=true
which Xvfb
Xvfb :99 -screen 0 1024x768x24 > /dev/null 2>&1 &
sleep 3
set +x
exec "$@"

Or alternatively, start it using the built in pyvista.start_xvfb(). Be sure to install xvfb and libgl1-mesa-glx with:

sudo apt-get install libgl1-mesa-dev xvfb

Or using the package manager used by your environment.